Hi,
is there anybody who already had a look into the KSMBD implementation. It is new in Kernel 5.15 and is the implementation of an SMBv3 Server on Linux Kernel side.
Just looking for some experience.
KSMBD - Kernel implementation of SMB Server
-
Just looking for some experience.
What are you hoping to accomplish? It doesn't do anything that samba doesn't. It is just faster (probably) while being less configurable. Unless you have some very fast networks, I don't see much reason to use it. And since it would require the proxmox kernel or a custom kernel ( no 5.15 in backports yet), it wouldn't make since to have a plugin yet. Most arm boards don't have a 5.15 kernel either.
-
Would it be a good, in the future, for ARM devices who are notoriously slow with SMB?
-
Would it be a good, in the future, for ARM devices who are notoriously slow with SMB?
Maybe but only if you are willing to give up a lot of features samba has.
-
ARM devices who are notoriously slow with SMB
I'm surprised by this statement. There is nothing like a typical ARM device!
Each hardware designer selects functions blocks and associated performance characteristics for a specific use-case.
Many cheap ARM devices are re-purposed mobile phone processors without the need for even low IO performance.
A Raspberry Pi 3 has a shared USB 2 bus (max capacity 480 MBits/s)
A Raspberry Pi 4 using BCM2711 processor was build for medium IO performance by using a PCIe x1 IO bus (max capacity 2,4 GBits/s).
Amazon uses purposed build ARM processors with high IO performance, as the result ARM systems are beating X86-64 systems for cloud computing workloads.
-
What are you hoping to accomplish? It doesn't do anything that samba doesn't. It is just faster (probably) while being less configurable. Unless you have some very fast networks, I don't see much reason to use it. And since it would require the proxmox kernel or a custom kernel ( no 5.15 in backports yet), it wouldn't make since to have a plugin yet. Most arm boards don't have a 5.15 kernel either.
From my perspective it would be interesting if the kernel implementation can have more throughput on old x86 hardware as Samba have it nowadays.
-
From my perspective it would be interesting if the kernel implementation can have more throughput on old x86 hardware as Samba have it nowadays.
How old? Unless you have more than GBe networking, it shouldn't make a difference. Most fairly recent x86 hardware can saturate gigabit and wouldn't need this.
Jetzt mitmachen!
Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!